31 January 2022 in IEEE Access
Deep learning has recently been shown to be effective in uncovering hidden patterns in non-Euclidean space, where data is represented as graphs with complex object relationships and interdependencies. Because of the implicit data dependence in the big graphs with millions of nodes and billions of edges, it is hard for industrial communities to exploit these methods to address real-world challenges at scale. The skewness property of big graphs, distributed file system performance penalty on small k-hop neighborhood subgraphs, and varying size of subgraph makes Graph Neural Networks (GNNs) training further challenging in a distributed environment using parameter servers. To address such issues, we propose a scalable, layered, fault-tolerance, and in-memory distributed computing-based graph neural network framework called Graph Distributed Learning Library (GDLL). The base layer utilizes an optimized distributed file system and a scalable graph data store to reduce the performance penalty. The second layer provides distributed graph processing using in-memory graph programming models while optimizing and hiding the underlying complexity of information complete subgraph computation. In the third layer, GNN modules are deployed on top of the first two layers for efficient distributed training using parameter servers. Finally, we evaluate and compare GDLL with the state-of-the-art solutions and outperform it significantly in terms of efficiency while maintaining similar GNN convergence.
4 March 2025 in IEEE Access
The Fitzpatrick scale is a widely used tool in dermatology for categorizing skin types based on melanin levels and sensitivity to ultraviolet light. The primary objective of this study is to enhance the accuracy of Fitzpatrick scale classification by addressing limitations in existing methodologies. Current approaches either rely on custom-designed hardware or utilize the Individual Typology Angle for image classification; however, these methods typically allow for a one-tone difference in classification and achieve a maximum accuracy of approximately 75%. A primary task for the skin tone classification is to apply filters to detect skin regions in an image, but the filters proposed for detecting skin do not apply to general datasets. In this paper, we propose a novel classification method that employs specialized filters to accurately detect and remove skin surface attributes, such as wrinkles and pores, using a controlled environment dataset obtained from a professional skin analyzer device. Our method involves modeling image features as a 9-dimensional feature vector, followed by a dimensionality reduction process to identify the most influential features and dominant areas within the feature space, thereby enabling deployment on low-power devices. We conducted extensive classification experiments using various Machine Learning algorithms. The results of our cross-validation tests demonstrate a significant improvement in classification accuracy, reaching up to 97%, thereby outperforming state-of-the-art methods without relaxing the accuracy criteria.
3 February 2026 in IEEE Transactions on Parallel and Distributed Systems
Since the emergence of Graph Neural Networks (GNNs), researchers have extensively investigated training on large-scale GNN training because of their success and wide usage in various domains including biological networks, finance, and recommendation systems. This work focuses on training large-scale distributed GNNs, where partitioning massive graphs across multiple machines creates remote communication overhead that becomes a major scalability bottleneck. We introduce a policy-driven caching mechanism that prioritizes node features and embeddings based on access frequency and cross-partition fetch cost, significantly minimizing communication overhead without sacrificing accuracy. Our policies are based on analysis of Node Affinities (NAFs) during multi-hop neighborhood sampling that extend substantially beyond the graph partition boundaries. Analyzing NAFs not only alleviates the communication bottleneck but also provides a systematic mechanism to manage in-memory data effectively, prioritizing GPU storage for node features with high fetch latency. We present FLARE, a system designed to handle partitioned feature data while leveraging the NAF-based caching policy. FLARE substantially reduces both communication overhead and training convergence time. Extensive experiments on benchmark datasets show that training FLARE on a three-layer GCN, GAT, and GraphSAGE across eight GPU machines achieves up to 12.04× (8.12× on average) speedup over DistDGLv2, demonstrating substantial performance gains compared to state-of-the-art methods.
1 January 2026 in IEEE Access